Spam-fighting Strategy
summary
Facebook describes an implemented pipeline that combines an online counter system and an offline clustering system to catch spam by aggregating weak signals into strong decisions. Counters use two-tally time-windowed metrics (e.g., total vs cold-opened URL posts) and binomial tests to flag anomalies; clusters store full member metadata (e.g., IPs, registration week, photo hashes) to retroactively remediate. Systems named include Sigma (rule engine), URL Classifier, and photo clustering; engineers report simple code, wide coverage, and practical success in production.
AI-written summary · about
chapters (7)
- 0:00Intro
- 1:00What is spam
- 1:57Aggregation
- 3:53Vocabulary
- 10:47Binomial Distribution
- 23:14Conclusion
- 25:34Questions
related talks
29mSocial Media Botnet Detection with Linkage Analysis and Machine Learning
35mDisinformation or Debate? Finding Signal Amid Political Noise
01h 00mRock ‘Em Sock ‘Em Robots: Bot Swatting Like the Pros
42mBuilding a Robust Fraud Detection System
59mCalculating by IP: How to Fight Spam on a Social Network
32mTricking the Trolls: A Look at the Digital Future
36mDetecting Influence Campaigns Across Social Media Platforms
25mTesting Your Organization's Social Media Awareness